Young’s modulus and residual stress of GeSbTephase-change thin films

Hammad Nazeera, Harish Bhaskaranb, Leon A Wolderinga, LeonAbelmanna,c,∗

aMESA+ Research Institute, University of TwentebDepartment of Materials, University of Oxford, Oxford, United Kingdom

cKIST Europe, Saarbrucken, Germany

Abstract

The mechanical properties of phase change materials alter when the phase is

transformed. In this paper, we report on experiments that determine the change

in crucial parameters such as Young’s modulus and residual stress for two of

the most widely employed compositions of phase change films, Ge1Sb2Te4 and

Ge2Sb2Te5, using an accurate microcantilever methodology. The results support

understanding of the exact mechanisms that account for the phase transition,

especially with regards to stress, which leads to drift in non-volatile data stor-

age. Moreover, detailed information on the change in mechanical properties will

enable the design of novel low-power nonvolatile MEMS.

Keywords: Youngs modulus, strain, phase change, GeSbTe, cantilever

resonance

1. Introduction

The present quest in Nanoelectromechanical Systems (NEMS) and Micro-

electromechanical Systems (MEMS) is for ever-lighter and lower power tech-

nologies, thus there is a constant effort to reduce device dimensions. The use

of phase change materials in such devices is a rather nascent concept, and is5

stymied by the lack of understanding of the exact nature of the mechanical

∗Corresponding authorEmail address: +49 6819382229, l.abelmann@utwente.nl (Leon Abelmann)

Preprint submitted to Thin Solid Films September 7, 2015

property change in such materials. However, an understanding of the phase

change transition will be transformational. For example, tunable resonators all

require high speed tuning of the resonance frequency of mechanical resonators.

This is mostly achieved using electrostatic tuning, which has other issues such10

as the requirement to maintain a voltage, electrostatic spring softening, which

competes with resonator stiffening due to tension in beam structures as well

as exacting impedance matching requirements for high-speed operation. The

modulation of a resonator material’s Young’s modulus is the most direct way

to affect the resonator’s frequency. This has been used in the reverse to detect15

magnetic fields [1, 2, 3], whereby the the magnetostriction in films is modulated

by magnetic fields, which in turns changes the Young’s Modulus of the micro-

cantilever. However, the use of phase change materials in such devices could be

potentially transformational, as these materials retain the phase state (amor-

phous/crystalline ratio) for many decades, thus requires infrequent tuning, but20

remain capable of a broad range depending on the change in modulus.

Phase change materials have also formed the dominant non-magnetic data

storage medium (from rewritable DVDs to Blu-Ray Disks) of the last couple

of decades, and more recently in the form of the highly scalable phase change

memories. In the context of the latter, where multiple bits are stored in a single25

cell, based on changes in resistivity, the stress induced by the phase change

itself could result in resistance drift [4]. However, a clear understanding of

these issues in such materials is strikingly absent. In the context of emerging

optoelectronics applications of these materials [5, 6, 7], such an understanding

would be vital to be able to design for variation induced by the change in the30

mechanical properties of these materials. This is especially so when coupled to

elements mismatched in thermal expansion coefficient, such as silicon-nitride.

Mechanical properties of thin films, notably their Young’s modulus, can

be accurately determined by means of microcantilevers [8, 9, 10, 11, 12, 13,

14]. The method is particularly usefull if the available substrate area is in the35

millimeter range, for instance due to small deposition area (such as in pulsed

laser deposition) or because the film has variations in mechanical properties on

2

the millimeter range. In these situations, methods like wafer curvature or X-ray

diffraction are difficult to apply, and less accurate.

The microcantilever method relies on a change in resonance frequency caused40

by the deposition of the thin film on the cantilever. The extra mass of the thin

film will decrease the resonance frequency, whereas the added rigidity of the

structure by adding the film will cause an increase. Up to now, thin films from

different deposition runs have been investigated, for instance to determine the

variation of the Young’s modulus with film composition [14]. In this paper45

we illustrate the method on GeSbTe (GST) phase change films, which has the

advantage that the Young’s modulus of the film can be changed when it is on the

cantilever, without changing its mass. The sole reason for a change in resonance

frequency will therefore be the change in the Young’s modulus of the phase

change film.50

Phase change materials based on the Ge-Sb-Te (GST) alloy are found to

exhibit excellent electrical and phase change properties and can endure large

numbers of read write cycles [15]. From the family of GST alloys, the com-

positions Ge1Sb2Te4 (GST124) and Ge2Sb2Te5 (GST225) are widely studied

because of their superior combination of properties [16, 17, 18]. These materi-55

als were mostly investigated on the basis of their optical, electrical and phase

change properties.

To support the use of these materials as an active device layer, information is

needed on the mechanical properties, like Young’s modulus and residual stress.

These properties have not been extensively investigated and a large variation60

of reported values can be found in literature [19, 20]. Knowledge on the com-

positional dependence of these properties and their variation with temperature

facilitates the choice for the correct composition for a particular application.

In this paper we report on accurate measurements of Young’s modululs and

residual stresses on technologically relevant thin films of phase change materi-65

als. The theoretical relations for the properties investigated in this study are

explained in section 2. Cantilevers were fabricated with a standard MEMS

fabrication process and GST thin films were sputtered on these cantilevers (see

3

500 mm

Cantilevers

Handle wafer

L

Si

Si

SiO2

GSTtf ts

ξ

Figure 1: Top: Scanning electron micrograph of cantilevers used, fabricated from a 3 µm

thick silicon device layer. Their length varies from 250 µm to 350 µm in steps of 10 µm. The

cantilevers have a constant width of 30 µm. Bottom: Model used for calculation.

section 3). In this section, we explain the annealing treatment and measurement

procedures as well. In addition to the determination of the Young’s modulus,70

we determined the residual stress of the GST thin films from the change in

static deflection of the cantilevers. Since crystallization of the GST thin films

is accompanied by a volume reduction [21], it results in an abrupt change in

the cantilever deflection. The mechanical measurements are complemented by

electrical characterisation and discussed in section 4.75

2. Theory

The resonance frequency shift and bending due to crystallisation of the

phase-change material can be calculated by assuming a simple, single sided

clamped cantilever (Figure 1).

2.1. Analytical model for the Young’s modulus of GST thin films in amorphous80

and crystalline states

A dynamic approach, based on the Euler-Bernoulli beam equation and the

dependence of the resonance frequency on the flexural rigidity of composite

cantilever, was used to develop an analytical model for determination of the

Young’s modulus [11]. In contrast to the work of Won et al, no approximation85

has been made [13].

4

Due to the addition of a thin film on the cantilever, the flexural rigidity of

the composite cantilever increases and its neutral plane (zero strain) shifts. In

addition the film will add to the mass of the composite cantilever. These three

effects result in a change in the resonance frequency. By measuring this change90

before and after deposition of the thin film, and before and after crystallisation,

one can obtain the variation in Young’s modulus from equation 1.

E∗f = 1

t3f

[6(γs + γf)B − 2E∗

s t3s − 3tfE

∗s t

2s − 2E∗

s tst2f

+2

√√√√√√√√E∗2

s t2s t4f + 3E∗2

s t3s t3f + (4E∗2

s t4s − 3AB)t2f +

(3E∗2s t5s − 9ABts)tf + E∗2

s t6s − 6ABt2s+

9(γs + γf)2B2

],

(1)

where

A = E∗s ts(γs + γf),

and

B = (

√E∗

s t3s

12tsρs− 0.568π∆f0L

2)2.

In the above equation L, t and E∗ are the length [m], thickness [m] and95

the effective Young’s modulus [Pa] respectively. The effective density γ = ρt

[kg/m2]. Subscripts ‘s’ and ‘f’ denote the silicon and the thin film. The measured

difference in the fundamental resonance frequency of the cantilevers before and

after the deposition of the thin film is denoted by ∆f0. By taking this difference,

any potential uncertainties in the thickness of the cantilever can be eliminated100

and a more accurate result is obtained [22].

The resonance frequency of the cantilever is a function of the Young’s mod-

ulus of the GST film and its mass, which is the product of density, density and

area. We assume the mass and area of the film do not change during a transition

from the amorphous to crystalline phase. The change in resonance frequency105

therefore is in first order only caused by a change in Young’s modulus. Due

5

to crystalisation however, the density increases, resulting in a decrease in film

thickness and a shift of the neutral plane towards the cantilever axis. This will

induce a resonance frequency shift by itself. Fortunately, this is a negligible

effect. From equation 1 we calculate that for a 200 nm GST film deposited on a110

3 µm silicon cantilever, the product of Young’s modulus and film thickness only

changes by 0.3% when reducing the film thickness by 6.5%. Therefore, we con-

sider the product of GST Young’s modulus and film thickness to be independent

on a reduction in film thickness due to the crystallization process.

2.2. Residual stress115

Crystallization of the GST thin films at elevated temperature leads to a re-

duction in volume. When deposited on cantilevers, this reduction leads to stress

which causes the cantilevers to bend upwards. By measuring the static deflec-

tion of these cantilevers before and after the deposition at every annealing step,

we can determine the residual stress in GST thin films at different temperatures.120

Since the GST film thickness is small compared to the substrate thickness, we

can use Stoney’s approximation [23, 24].

σf =1

3

Est2s ξ

(1 − νs)tfL2, (2)

the symbols σ, E, ν, t, L, and ξ are the residual stress, Young’s modulus,

Poison ratio, thickness, length and deflection respectively. Subscripts ‘s’ and ‘f’

denote the silicon and thin film. Again, the product of residual stress and film125

thickness is independent on the film thickness itself.

3. Experimental details

3.1. Fabrication of cantilevers

The cantilever fabrication process given in the appendix. It is an improved

version of the process previously reported in [11]. The main differences are that130

we used a different recipe for the back side etch, which enables us to use foil on

the front side instead of polyimide pyralin. After etching, the foil on the front

6

side and photoresist material from the back side of the wafer were removed using

O2 plasma (Tepla300E). Subsequently, the cantilevers were released by etching

the BOX layer using buffered-hydrofluoric acid.135

Scanning electron micrographs (SEM) as shown in figure 1 were used to

inspect and characterise the fabricated cantilevers.

3.2. GST deposition

The 200 nm films of compositions Ge1Sb2Te4 (GST124) and Ge2Sb2Te5

(GST225) were deposited directly on the Si cantilevers by DC magnetron sput-140

tering in an argon plasma at a sputtering power of 300 W and deposition rate of

6 nm/s. A 5 nm ZnS-SiO2 capping layer was deposited to protect the films from

oxidation at a sputtering power of 1 kW at a rate of 3.4 nm/s. Measurement of

film thicknesses on microcantilevers is difficult and inaccurate. Therefore films

thicknesses are estimated from deposition rates, calibrated by low angle x-ray145

diffraction and TEM measurements.

3.3. Annealing of GST

Compositional and phase dependence of the Young’s modulus, residual stress

and sheet resistance of the GST225 and GST124 thin films were investigated

at different temperatures. Samples were annealed from room temperature to150

the desired temperature with a ramping rate of 3C/min in a vacuum furnace

with nitrogen environment at a pressure of 1 mbar. The chips were kept at the

desired temperature for 15 minutes and then cooled down to room temperature.

The cooling rate was not controlled, but much lower than the heating rate.

Measurements for the Young’s modulus, residual stress and sheet resistance155

were conducted at room temperature after each annealing step. Samples of

different composition, identified by GST225 and GST124, were passed through

an identical annealing procedure up to 180C with reduced temperature steps

around 150C. In order to investigate the behaviour of the properties around the

crystallization temperature of GST124, a third sample with a Ge1Sb2Te4 com-160

position (which we will refer to as GST124∗) was annealed with extra temper-

7

ature steps around 130C. The annealing history of the samples is summarized

in table 1.

It should be noted that the cantilever fabrication process includes temper-

atures in excess of 350 C, far above the temperatures used for the annealing165

of the GST films. Changes in cantilever resonance frequency and bending can

therefore solely be attributed to the changes in the GST films.

Table 1: Annealing history of the GST thin films. GST124∗ passed through more annealing

steps before reaching 140C.

Annealing temperature (C)

Sample 60 100 110 120 130 140 150 160 170 180

GST225 x x x x x x x

GST124 x x x x x x x

GST124∗ x x x x x x x x x x

3.4. Resonance frequency measurements

The resonance frequency of the cantilevers was measured by using a MSA-

400 micro system analyzer scanning laser Doppler vibrometer. Measurements170

of the resonance frequency were conducted both before and after the deposition

and each annealing step. The thermal vibration at room temperature of the

cantilevers in ambient conditions was used to measure the amplitude spectrum,

which is then fitted to the theoretical expression of mass-spring system with

damping to estimate the resonance frequency.175

3.5. Static deflection measurements

Static deflection ξ of the cantilevers was measured at room temperature by

using white light interferometry (WLIF) of a Polytec MSA-400 analyzer. From

the deflection, the residual stress in the GST thin films was determined using

equation 2. As expected, all cantilevers were found to be straight before de-180

position of the GST thin films. After annealing, all cantilevers bent upwards,

8

0 100 200

-

-

-

0

15

30y (m

m)

50 150 250

l (mm)0

2

4

8

6--

---

z (m

m)

-50

CantileverThrough holeAnchor

Figure 2: Top view of the reconstructed image from a white light interference measurement

of an approximately 290 µm long, 30 µm wide and 3 µm thick silicon cantilever, with 200 nm

GST225 thin film, after annealing at 150C.

indicating that the GST films develop a tensile stress. A 2-D white light in-

terference microscopic image of a 250 µm long cantilever with GST225 after

annealing at 150C is shown in figure 2. From this image, the profile of the can-

tilever is obtained by averaging over the width. Cantilevers of varying length185

were measured, to reduce the error in the measurement due to uncertainty in

the location of the cantilever base. All static deflection measurements were

conducted at room temperature.

3.6. Sheet resistance measurements

The sheet resistance of the GST thin films was measured by a four-point190

probe on the handle wafer. The measurements were taken after the deposition

of the GST thin films and after each annealing step at room temperature.

4. Results and Discussion

After deposition of the GST layers, the resonance frequencies of the can-

tilevers decrease by about 2 kHz and they remain essentially straight. Since195

after deposition the GST layers are amorphous, they can be easily changed to

the crystalline state by annealing in a standard oven. After anneal, the reso-

nance frequency increases by about 500 Hz (figure 3), and the cantilevers bend

upwards. Figure 4 shows that the end of a 250 µm long cantilever bends up

as much as 6 µm. These considerable effects can be used to accurately deter-200

9

mine the changes in Youngs modulus, stress and strain of the different GST

compositions.

4.1. Young’s modulus

From the shift in resonance frequency, the in-plane Young’s modulus can be

determined (see section 2). The Young modules of the GST thin films is shown205

in figure 5 as a function of composition and annealing step. In the amorphous

state, below the crystallization temperature, the Young’s modulus is lower than

that of the corresponding crystalline phase [25].

During crystallization the thickness of the GST reduces, leading to an equal

increase in film density. Since the mass and area of the film remain constant, the210

product of thickness and density will remains constant as well. The same is true,

to a good approximation, for the product of Young’s modulus and thickness.

To be able to give absolute values, we have assumed the thickness reduction to

be 6.5% for the GST225 composition [26, 27] (leading to an increase in density

from 5870±50 kg/m3 to 6270±20 kg/m3), and a reduction of 4% for GST124 [28]215

(with density increasing from 5900 kg/m3 to 6200 kg/m3).

The measured Young’s modulus of the GST225 thin film as deposited in

amorphous state was found to be 18.9 GPa with a standard error of 0.7 GPa

(indicate in the following by (0.7)). The Young’s modulus increased sharply

above the crystallization temperature of 150C to a value of 38.2 (0.3) GPa.220

The crystallization temperature agrees well with the range of values quoted in

literature [16, 26, 29, 30]. The increase in the Young’s modulus from amorphous

to crystalline state is consistent with the results published in literature [31, 32,

13]. We assumed a typical value of νf of 0.3, in order to compare with the biaxial

modulus values published in literature.225

The Young’s modulus of the GST124 thin film was found to be 15.9 (0.2) GPa

and 31.3 (0.3) GPa in the amorphous and crystalline state respectively. The

crystallization temperature of 130C is consistent with the range of tempera-

ture values reported elsewhere [28, 33]. The fact that the crystallization tem-

perature of GST124 is lower than GST225 agrees with measurements by Car-230

10

39 42 45 48 51

0.0

0.5

1.0

1.5Ge

2Sb

2Te

5

Si cantilever

GST amorphous

GST crystalline

Nor

mal

ised

am

plitu

de

Frequency, fo

(kHz)

54 57 60 63 66

0.0

0.5

1.0

1.5

Nor

mal

ised

am

plitu

de

Frequency, fo(kHz)

Ge1Sb

2Te

4

*Si cantilever

GST amorphous

GST crystalline

Figure 3: Measured resonance frequencies of the silicon cantilever without (black) and with

GST thin films in amorphous (red) and crystalline state (blue). The resonance frequency

decreases after deposition, whereas it increases upon annealing. This increase is attributed to

the higher Young’s modulus of the GST thin films in crystalline state. Top: ( Ge2Sb2Te5 on

a ∼ 310 µm long cantilever. Bottom: Ge1Sb2Te4 on a ∼ 260 µm long cantilever.

11

µ

µ

ξ

Figure 4: Cantilever profile after annealing, obtained from white light interferometry (figure 2).

The measurements are taken at room temperature. The maximum upward static deflection

increases with annealing temperature because of the built up of tensile residual stress.

12

Figure 5: Young’s modulus of the GST225 and GST124 thin films plotted as a function of

the annealing temperature. We observe a rise in the Young’s modulus value above the crys-

tallization temperature for both the GST225 and GST124 thin films. The Young’s modulus

of the GST225 film is 30% higher than that of GST124. The GST124 composition marked

with an asterix has a different anneal sequence, see table1. The lines are guides to the eye.

ria [15] and Kalb [34]. The measured Young’s moduli are lower than reported

by Blachowicz et al. [28], who found 24.8 (0.06) GPa for the amorphous and

39.5 (0.8) GPa for the crystalline phase. In this work the Youngs modulus is

derived from Brillouin light scattering experiments at phonon frequencies in the

range of 2 to 16 GHz, so far above the resonance frequency of our cantilevers.235

Since the Youngs modulus in rigid materials is expected to increase slightly with

frequency [35], this could very well explain the difference between both methods.

4.2. Residual stress

From the cantilever bending after annealing, the residual stress in the GST

layers can be determined. This stress strongly depends on the annealing tem-240

perature as well as the material composition (see figure 6). We observed a

13

sharp increase in the residual stress, when the films were annealed above the

crystallization temperature.

There is no residual stress in the as-deposited GST225 thin film. This is

agreement with work by Leervad-Pedersen et al [31], who demonstrate that245

stress is released in the amporphous state due to plastic flow. The stress in-

creases slowly to a value of 49.7 (0.3) MPa at 140C, just before the crystalliza-

tion temperature. A sharp increase in residual stress to a value of 331.6 (1.5) MPa

was observed after annealing at 150C. The crystallization temperature deter-

mined from the stress measurements is consistent with the Young’s modulus250

results. Annealing above the crystallization temperature up to 180C shows

some stress relaxation, see figure 6. The trend in the temperature dependence

of the residual stress in GST225 agrees with results published by others [31, 36].

Krusin-Elbaum et al. observed a similar stress release, and showed it to be

dependant on the material composition [29].255

Residual stress in the GST124 and GST124∗ thin films were also found to

be negligible in the as-deposited state. The stress for GST124∗ increased to

a value of 46.8 MPa (0.2) at 120C. When measured after annealing at the

crystallization temperature (130C), the residual stress increased to value of

238.6 (1.0) MPa. The GST124 and GST124∗ sample show distinct differences260

in residual stress after annealing above the crystallization temperature. The an-

nealing history has a strong effect. The GST124∗ sample, which passed through

more annealing steps (see table 1) shows a 45% higher residual stress than the

GST124 sample at 140C. It is known that the crystallisation process is a func-

tion of the temperature as well as time [37]. As the temperature ramp rate265

increases, the grainsize in the crystalline film decreases. It is not unlikely that

this has a profound effect on the residual stress.

Unlike the GST225 thin film however, stress relaxation in the GST124 thin

films at elevated annealing temperatures is not present. We suspect that this is

related to the fact that in GST225 has a meta-stable FCC crystal phase [26],270

which might be absent in GST124.

14

Figure 6: Residual stress dependence of Ge2Sb2Te5 and Ge1Sb2Te4 films on temperature.

The Ge1Sb2Te4 films shows lower values of residual stress as compared to Ge2Sb2Te5. There

is a clear difference in residual stress values for two identical Ge1Sb2Te4 samples, which were

annealed through different steps (see table 1). Stress relaxation as seen in the Ge2Sb2Te5 film

after 150C was not observed for Ge1Sb2Te4 films. The lines are guides to the eye.

15

4.3. Crystallization temperature

The variation in the Young’s modulus and residual stress of the GST225

thin films with annealing temperature is compared with changes in the trans-

missivity data obtained from literature [38] in figure 7. The observed dip in the275

transimissivity coincides well with the rise of the Young’s modulus and resid-

ual stress. This crystallization temperature of 150C film agrees as well with

data published for Ge2Sb2Te5 in [32, 39]. Therefore the observed changes in

mechancal properties have the same origin as the change in transmissivity, and

are therefore caused by crystallisation.280

4.4. Strain

From the measured residual stress and the Young’s modulus at room tem-

perature, we can calculate the strain in the film. The result is plotted in figure 8.

The drop in the residual stress of the GST225 thin film after the crystallization

temperature (figure 6) is also reflected in a steep drop in the strain values. Like-285

wise, the strain in the GST124 thin films follows the stress behaviour after the

crystallization temperature, because of the negligible variation in the Young’s

modulus.

4.5. Sheet resistance

Next to change in Youngs modules, increase in stress and reduction of trans-290

missivity, the crystallisation of the GST layers also leads to a change in electrical

resistance. Therefore the sheet resistance was measured as a function of the an-

nealing temperature, see figure 9. The sheet resistance of both the GST225 and

GST124 thin films in the as-deposited amorphous state is high (in the range of

1100 to 1200 kΩ/sq) compared to the crystalline state. Annealing with increas-295

ing temperature reduces the sheet resistance monotonously. This is in contrast

with the sudden changes in mechanical properties, and with previously reported

work. Siegrist [40], Jang et al. [41] and Njoroge et al. [26] for instance observed

a sudden drop in resistance around the annealing temperature.

16

Figure 7: Young’s modulus and residual stress of the GST225 thin film compared to the

change in transmissivity at the crystallization temperature [38]. The lines are guides to the

eye.

17

Figure 8: Strain in the GST thin films as a function of annealing temperature. The strain

is estimated from the measured Young’s modulus and residual stress at the corresponding

annealing temperature.

18

Ω

Figure 9: Change in the sheet resistance of the GST225 and GST124 films as a function of tem-

perature. The sheet resistance was measured using a four-point probe at room temperature.

The lines are guides to the eye.

At a the crystallisation temperature of 150C, the sheet resistance was300

74 kΩ/sq for the GST225 thin film, which is in the same order of magnitude as

reported in literature. However, the sheet resistance in amorphous state is one

order of magnitude less than reported values in literature (∼ 107 ohm/sq). We

suspect that these lower values, and the absence of a sudden drop, are caused

by the silicon device layer underneath the GST thin films. A shortcut current305

through this layer might reduce the sheets resistance and obscure a sudden drop

at the crystallisation temperature. The effect of the resistance measurement it-

self on the phase change transition should however also be considered [42].

The sheet resistance of the GST124 and GST124∗ samples are generally

higher than the GST225 film. At 140C was found to be 184 kΩ/sq and310

164 kΩ/sq respectively (see figure 9). Above 150C however, we again observe

a clear dependence on annealing history.

19

5. Conclusion

We have deposited amorphous 200 nm GST films on 3 µm thick, 250 to

350 µm long silicon cantilevers, and heated them above the crystallisation tem-315

perature to induce a phase change. After annealing, the cantilever resonance

frequency shifts up by approximately 500 Hz and the cantilevers bend upwards

by about 6 µm.

From these changes in resonance frequency and radius of bending curvature

we can determine the change in Young’s modulus and residual stress when pass-320

ing from the amorphous to crystalline state. Calculations show that the effect

of a change in neutral line when the thickness of the GST film changes by 6.5%

is negligble. Therefore, we can consider the product of GST Young’s modulus

and film thickness, as well as the product of residual stress and film thickness,

to be independent on a reduction in film thickness due to the crystalllization325

process.

We investigate two compositions, Ge1Sb2Te4 (GST124) and Ge2Sb2Te5 (GST225).

The Young’s modulus increases sharply from 18.9 (0.7) GPa to 38.2 (0.3) GPa

(GST225) and 15.9 (0.2) GPa to 31.3 (0.3) GPa (GST124) after annealing above

the crystallization temperature. The crystallization temperature of GST225330

(150C) was slightly higher than that of GST124 (130C). Both values agree

well with values quoted in literature obtained by optical reflection [38] and elec-

trical conductivity [26].

Residual stress in the GST thin films increases sharply from almost no stress

after deposition to values of 331.6 (1.5) MPa (GST225) and 238.6 (1.0) MPa335

(GST124∗) when changing from the amorphous to crystalline phase. The in-

crease of stress follows the same temperature behaviour as measured for the

Young’s modulus.

We observed relaxation in the residual stress of the GST225 thin film when

annealed above the crystallization temperature. This relaxation is not present340

in the GST124 films. The residual stress is highly dependent on the annealing

history, we observed higher stress values if the film is annealed longer below the

20

crystallization temperature.

The sheet resistance measured for the two compositions of GST shows one

order of magnitude difference between the amorphous and crystalline state. Un-345

like the Young’s modulus and residual stress, there is no sharp transition tem-

perature. The resistance rather drops monotonously over a wide temperature

range. This might be caused by the fact that we measure the sheet resistance

of the phase change material directly on the semiconducting silicon substrate.

The cantilever based method analysis method presented here provides valu-350

able information on the mechanical properties of GST124 and GST224 phase

change films. It can be easily extended to films of different thicknesses and

composition. Moreover, we demonstrate that phase change materials allow for

significant changes in resonance frequency and curvature of micro-cantilevers.

Since crystallisation of phase change films is in principle reversible, we believe355

they can be succesfully applied as actuator material in low-power micromechan-

ical devices. We envision tuneable resonantors that do not require energy after

tuning the resonance frequencies, or switches that only require power during a

change in position.

Acknowledgements360

The authors are indebted to Andrew Pauza of Plarion inc. for the deposi-

tion of the GST film, Meint de Boer for etching, Remco (Pino) Sanders for laser

Doppler vibrometer measurements, Johnny Sanderink and Henk van Wolferen

for SEM. The authors would like to thank dr. Niels Tas and prof. Miko Elwen-

spoek of the University of Twente, prof. Matthias Wuttig of the RWTH Aachen365

and prof. David Wright of Exeter University for fruitful discussion.

The authors gratefully acknowledge the support of the SmartMix Program

(SmartPie) of the Netherlands Ministry of Economic Affairs and the Netherlands

Ministry of Education, Culture and Science. HB is grateful to EPSRC for

funding via grants EP/J00541X/2 and EP/J018694/1370

21

References

[1] P. Zhao, Z. Zhao, D. Hunter, R. Suchoski, C. Gao, S. Mathews, M. Wuttig,

I. Takeuchi, Fabrication and characterization of all-thin-film magnetoelec-

tric sensors, Applied Physics Letters 94 (24) (2009) 243507.

[2] S. Dong, J. Zhai, F. Bai, J.-F. Li, D. Viehland, Push-pull mode magne-375

tostrictive/piezoelectric laminate composite with an enhanced magneto-

electric voltage coefficient, Applied Physics Letters 87 (6) (2005) 062502.

[3] H. Bhaskaran, M. Li, D. Garcia-Sanchez, P. Zhao, I. Takeuchi, H. X. Tang,

Active microcantilevers based on piezoresistive ferromagnetic thin films,

Appl. Phys. Lett. 98 (1) (2011) 013502.380

[4] A. Sebastian, N. Papandreou, A. Pantazi, H. Pozidis, E. Eleftheriou, Non-

resistance-based cell-state metric for phase-change memory, J. Appl. Phys.

110 (8) (2011) 084505.

[5] W. Pernice, H. Bhaskaran, Photonic non-volatile memories using phase

change materials, Applied Physics Letters 101 (17) (2012) 171101.385

[6] C. Rios, P. Hosseini, C. Wright, H. Bhaskaran, W. Pernice, On-chip pho-

tonic memory elements employing phase-change materials, Advanced Ma-

terials 26 (9) (2014) 1372–1377.

[7] P. Hosseini, C. Wright, H. Bhaskaran, An optoelectronic framework enabled

by low-dimensional phase-change films, Nature 511 (7508) (2014) 206–211.390

[8] R. H. Poelma, H. Sadeghian, S. P. M. Noijen, J. J. M. Zaal, G. Q. Zhang, A

numerical experimental approach for characterizing the elastic properties of

thin films: Application of nanocantilevers, J. Micromech. Microeng. 21 (6)

(2011) 065003.

[9] D. Isarakorn, A. Sambri, P. Janphuang, D. Briand, S. Gariglio, J.-M.395

Triscone, F. Guy, J. W. Reiner, C. H. Ahn, N. F. de Rooij, Epitaxial

22

piezoelectric MEMS on silicon, J. Micromech. Microeng. 20 (5) (2010) art.

no. 055008.

[10] H. Ræder, F. Tyholdt, W. Booij, F. Calame, N. P. Østbø, R. Bredesen,

K. Prume, G. Rijnders, P. Muralt, Taking piezoelectric microsystems from400

the laboratory to production, J. Electroceram. 19 (4) (2007) 357–362.

[11] H. Nazeer, M. D. Nguyen, L. A. Woldering, L. Abelmann, G. Rijnders,

M. C. Elwenspoek, Determination of the Young’s modulus of pulsed laser

deposited epitaxial PZT thin films, J. Micromech. Microeng. 21 (7) (2011)

074008.405

[12] H. Nazeer, L. A. Woldering, L. Abelmann, M. D. Nguyen, G. Rijnders,

M. C. Elwenspoek, Influence of silicon orientation and cantilever undercut

on the determination of the Young’s modulus of thin films, Microelectron.

Eng. 88 (8) (2011) 2345–2348.

[13] Y. Won, J. Lee, M. Asheghi, T. Kenny, K. Goodson, Phase and thickness410

dependent modulus of Ge2Sb2Te5 films down to 25 nm thickness, Applied

Physics Letters 100 (16) (2012) 161905.

[14] H. Nazeer, M. D. Nguyen, A. J. H. M. Rijnders, L. Abelmann, M. C. Elwen-

spoek, Compositional dependence of the young’s modulus and piezoelectric

coefficient of (110)-oriented pulsed laser deposited pzt thin films, Journal415

of microelectromechanical systems 24 (1) (2014) 166–173.

[15] E. Carria, A. M. Mio, S. Gibilisco, M. Miritello, M. G. Grimaldi, E. Rimini,

Tuning the crystallization temperature of amorphous Ge2Sb2Te5 by O and

Si recoil implantation, Electrochem. Solid-State Lett. 14 (3) (2011) H124–

H127.420

[16] J. Tomforde, W. Bensch, L. Kienle, V. Duppel, P. Merkelbach, M. Wuttig,

Thin films of Ge-Sb-Te-based phase change materials: Microstructure and

in situ transformation, Chem. Mater. 23 (17) (2011) 3871–3878.

23

[17] M. Krbal, A. V. Kolobov, J. Haines, A. Pradel, M. Ribes, P. Fons,

J. Tominaga, C. Levelut, R. Le Parc, M. Hanfland, Temperature inde-425

pendence of pressure-induced amorphization of the phase-change memory

alloy Ge2Sb2Te5, Appl. Phys. Lett. 93 (3) (2008) 031918.

[18] T. Matsunaga, N. Yamada, Structural investigation of GeSb2Te4: A

high-speed phase-change material, Phys. Rev. B 69 (10) (2004) 1041111–

1041118.430

[19] A. Marmier, K. Kohary, C. D. Wright, Determination of the anisotropic

elastic properties of Ge1Sb2Te4, Appl. Phys. Lett. 98 (23) (2011) 231911.

[20] C.-A. Jong, W. Fang, C.-M. Lee, T.-S. Chin, Mechanical properties of

phase-change recording media: GeSbTe films, Jpn. J. Appl. Phys. 40 (5 A)

(2001) 3320–3325.435

[21] V. Weidenhof, I. Friedrich, S. Ziegler, M. Wuttig, Atomic force microscopy

study of laser induced phase transitions in Ge2Sb2Te5, J. Appl. Phys.

86 (10) (1999) 5879–5887.

[22] J.-A. Schweitz, A new and simple micromechanical approach to the stress-

strain characterization of thin coatings, J. Micromech. Microeng. 1 (1)440

(1991) 10–15.

[23] G. Stoney, The tension of metallic films deposited by electrolysis, Proc. R.

Soc. Lond. A 82 (553) (1909) 172–175.

[24] S. Jeon, T. Thundat, Instant curvature measurement for microcantilever

sensors, Appl. Phys. Lett. 85 (6) (2004) 1083–1084.445

[25] D. Weaire, M. F. Ashby, J. Logan, M. J. Weins, On the use of pair potentials

to calculate the properties of amorphous metals, Acta Metallurgica 19 (8)

(1971) 779–788.

[26] W. K. Njoroge, H.-W. Woltgens, M. Wuttig, Density changes upon crys-

tallization of Ge2Sb2.04Te4.74 films, J. Vac. Sci. Technol. A 20 (1) (2002)450

230–233.

24

[27] S. Elliott, J. Hegedus, Computer simulation of the phase-change cycle of

GST-225, in: MRS Proceedings, 1072 , 1072-G01-02, Vol. 1072, San Fran-

cisco, CA, 2008, pp. 99–107.

[28] T. Blachowicz, M. G. Beghi, G. Guntherodt, B. Beschoten, H. Dieker,455

M. Wuttig, Crystalline phases in the GeSb2Te4 alloy system: Phase tran-

sitions and elastic properties, J. Appl. Phys. 102 (9) (2007) 093519.

[29] L. Krusin-Elbaum, C. Cabral, K. N. Chen, M. Copel, D. W. Abraham,

K. B. Reuter, S. M. Rossnagel, J. Bruley, V. R. Deline, Evidence for segre-

gation of Te in Ge2Sb2Te5 films: Effect on the ”phase-change” stress, Appl.460

Phys. Lett. 90 (14) (2007) 141902.

[30] J. Kalb, F. Spaepen, M. Wuttig, Calorimetric measurements of phase trans-

formations in thin films of amorphous Te alloys used for optical data stor-

age, J. Appl. Phys. 93 (5) (2003) 2389–2393.

[31] T. P. Leervad Pedersen, J. Kalb, W. K. Njoroge, D. Wamwangi, M. Wut-465

tig, F. Spaepen, Mechanical stresses upon crystallization in phase change

materials, Appl. Phys. Lett. 79 (22) (2001) 3597–3599.

[32] J. Kalb, F. Spaepen, T. P. Leervad Pedersen, M. Wuttig, Viscosity and

elastic constants of thin films of amorphous te alloys used for optical data

storage, J. Appl. Phys. 94 (8) (2003) 4908–4912.470

[33] B. Kalkan, S. Sen, B. G. Aitken, S. V. Raju, S. M. Clark, Negative P-T

slopes characterize phase change processes: Case of the Ge1Sb2Te4 phase

change alloy, Phys. Rev. B 84 (1) (2011) 014202.

[34] J. A. Kalb, M. Wuttig, F. Spaepen, Calorimetric measurements of struc-

tural relaxation and glass transition temperatures in sputtered films of475

amorphous Te alloys used for phase change recording, J. Mater. Res. 22 (3)

(2007) 748–754.

25

[35] T. Pritz, Frequency dependences of complex moduli and complex poisson’s

ratio of real solid materials, Journal of Sound and Vibration 214 (1) (1998)

83–104.480

[36] I.-M. Park, J.-K. Jung, S.-O. Ryu, K.-J. Choi, B.-G. Yu, Y.-B. Park, S. M.

Han, Y.-C. Joo, Thermomechanical properties and mechanical stresses of

Ge2Sb2Te5 films in phase-change random access memory, Thin Solid Films

517 (2) (2008) 848–852.

[37] G. Burr, P. Tchoulfian, T. Topuria, C. Nyffeler, K. Virwani, A. Padilla,485

R. Shelby, M. Eskandari, B. Jackson, B.-S. Lee, Observation and modeling

of polycrystalline grain formation in Ge2Sb2Te5, Journal of Applied Physics

111 (10) (2012) 104308.

[38] N. Yamada, E. Ohno, K. Nishiuchi, N. Akahira, M. Takao, Rapid-phase

transitions of GeTe-Sb2Te3 pseudobinary amorphous thin films for an op-490

tical disk memory, J. Appl. Phys. 69 (5) (1991) 2849–2856.

[39] I. Friedrich, V. Weidenhof, W. Njoroge, P. Franz, M. Wuttig, Structural

transformations of Ge2Sb2Te5 films studied by electrical resistance mea-

surements, J. Appl. Phys. 87 (9 I) (2000) 4130–4134.

[40] T. Siegrist, P. Jost, H. Volker, M. Woda, P. Merkelbach, C. Schlockermann,495

M. Wuttig, Disorder-induced localization in crystalline phase-change ma-

terials, Nature Materials 10 (3) (2011) 202–208.

[41] M. H. Jang, S. J. Park, D. H. Lim, M.-H. Cho, Y. K. Kim, H.-J.

Yi, H. S. Kim, Structural stability and phase-change characteristics of

Ge2Sb2Te5/SiO2 nano-multilayered films, Electrochem. Solid State Letters500

12 (4) (2009) H151–H154.

[42] Z.-Q. Cheng, G.-G.and Zhang, J.-N. Ding, Z. Ling, Y.-B. Chen, Mechanical

and electrical properties of gesb2te4 film with external voltage applied,

Applied Surface Science 285 (PARTB) (2013) 532–537.

26

[43] H. V. Jansen, M. J. De Boer, S. Unnikrishnan, M. C. Louwerse, M. C.505

Elwenspoek, Black silicon method X: a review on high speed and selective

plasma etching of silicon with profile control: an in-depth comparison be-

tween Bosch and cryostat DRIE processes as a roadmap to next generation

equipment, Journal of Micromechanics and Microengineering 19 (3) (2009)

033001.510

[44] R. A. Brookhuis, T. S. J. Lammerink, R. J. Wiegerink, M. J. De Boer, M. C.

Elwenspoek, Force sensor for measuring power transfer between the human

body and the environment, in: Proc. 16th International Solid-State Sen-

sors, Actuators and Microsystems Conference, TRANSDUCERS’11, Bei-

jing, 2011, pp. 2042–2045.515

6. Appendix

6.1. Fabrication process

Figure 10 shows the fabrication scheme for the cantilevers using a dedicated

SOI/MEMS fabrication process. The process is described briefly as follows.

(a) A double side polished silicon on insulator (SOI) wafer was selected. The520

substrate has a 380 µm handle wafer and a 3 µm thick device layer. The device

layer defines the thickness of the cantilevers. A layer of 500 nm buried oxide

(BOX) serves as an etch stop during the etching of the device layer and handle

wafer. A photoresist mask was designed with cantilevers that have varying

lengths from 250 µm to 350 µm in steps of 10 µm. The cantilevers have a fixed525

width of 30 µm. (b and c) The front side of the (001) single crystal silicon device

layers was patterned by conventional UV photolithography to define the shape

of the cantilevers. The photoresist used was Olin907-17 with a thickness of

1.7 µm. (d) Subsequently the cantilevers were anisotropically etched by deep

reactive ion etching (DRIE) using SF6, O2 and C4F8 gases [43]. (e and f) A530

layer of 3.5 µm thick photoresist (908-35) was applied on the back side of the

wafer and patterned to define holes for cantilevers release. (g) DuPont MX-5020

27

SiliconSiO2

PhotoresistDuPont foil

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)

Figure 10: Fabrication process of the silicon cantilevers. (a) A SOI wafer with a 3 µm thick

device layer and a 500 nm thick BOX was selected. (b and c) Application and patterning

of the photoresist (Olin 907-17) on the front side of the wafer. (d) Silicon device layer was

etched by DRIE. (e and f) Thick photoresist (Olin 908-35) was applied and patterned on the

back side of the wafer. (g) Application of the foil (DuPont MX5020) on the front side for

stable temperature control and avoid helium leakage. (h) Through holes from the back side

were etched by DRIE. (i) Photoresist and foil was removed from the front and back side using

oxygen plasma. (j) Cantilevers were released by etching the BOX layer using BHF.

foil was applied on the front side of the wafers to protect the cantilevers from

damage and prevent leakage of the helium during the wafer through etching.

Application of this foil was required to ensure stable temperature control of535

the wafer during the back side DRIE process [44]. (h) Etching of the handle

wafer from the back side was then performed by DRIE using SF6, O2 and C4F8

gases [43].

28